PitGuard

A mine safety dashboard that tracks gas, air quality, temperature and humidity across zones and raises alerts when readings look dangerous. It runs on simulated sensor data.

I hand-designed the MySQL schema (zones, workers, sensor readings, alerts, users) and wrote rule-based risk logic. For ML hazard detection I chose Isolation Forest, an unsupervised method, because any labels would only come from my own thresholds. Built with Flask, Flask-SQLAlchemy and MySQL, with an HTML, Tailwind and vanilla JS front end.

No measured result: the sensor data is simulated. Deployed on Render + Aiven.

With simulated data, any labels would only come from my own thresholds, so an unsupervised model (Isolation Forest) was the honest choice over a supervised one. I designed the schema and the risk logic by hand, and let AI help with the routine CRUD routes and styling.